论文部分内容阅读
推广性能是机器学习理论研究的主要目的之一。为了研究相依序列下采用ERM算法的学习机器的推广性能,本文基于β-混合序列建立了采用ERM算法的学习机器的经验风险到它的期望风险相对一致收敛速率的界。这个界不仅把基于独立序列下已有的结果推广到β-混合相依序列的情况,而且对β-混合相依序列现有的一些结论进行了改进。得到了β-混合相依序列下,采用ERM算法的学习机器的推广性能的界。
The promotion of performance is one of the main purposes of machine learning theory research. In order to study the generalized performance of learning machines using ERM algorithm in the dependent sequence, this paper establishes the bound of the relative uniform convergence rate of the expected risks from the empirical risk of the ERM-based learning machine to the β-mixed sequence. This field not only extends the existing results based on independent sequences to β-mixed dependent sequences, but also improves some of the existing conclusions of β-mixed dependent sequences. The bounds of the promotion performance of learning machines using ERM algorithm are obtained under the β-mixing dependent sequence.